search

wenwenyu / PICK-pytorch

Code for the paper "PICK: Processing Key Information Extraction from Documents using Improved Graph Learning-Convolutional Networks" (ICPR 2020)
https://arxiv.org/abs/2004.07464
MIT License
553 stars 191 forks source link

last drop out may affect the validation result #77

Closed jalola closed 3 years ago

jalola commented 3 years ago

Since the last drop out of the encoder is not created as an attribute of the encoder module, the validation process will run it as well with drop out prob 0.1

wenwenyu commented 3 years ago

Hi, thanks for your reminder. We already pass training args to the last dropout function of the encoder, so it actually didn't affect the validation result. But to be consistent with my usage of other dropout, I have modified the call method here. Thanks again.